Abstract

This paper reviews the current progress and outlook of various applications of machine learning techniques in the context of hydrogen fuel cell electric vehicles. Emphasizing the importance of efficient energy management systems, it presents a detailed comparative analysis of recent developments. Key research works on machine learning-based energy management systems are discussed, including the use of the Twin Delayed Deep Deterministic Policy Gradient algorithm, which has demonstrated up to 28% reduction in overall operational costs. Moreover, the review includes machine learning -based energy management systems approaches that account for fuel cell degradation, an area receiving growing attention due to its impact on performance and longevity. Notably, reinforcement learning strategies have achieved improvements in fuel economy of 5.7% and reductions in fuel cell degradation rates by 4.5%. In addition, robust machine learning -based prediction models are highlighted for their effectiveness in data-driven fault diagnosis, contributing to a 13.9% reduction in fuel usage during the Federal Test Procedure 75 Driving Cycle and a 14.32% reduction in the New European Driving Cycle, along with carbon dioxide emissions cut to less than 26.4%. Despite this progress, existing review papers seldom address the use of machine learning for predictive maintenance in the automotive sector, nor do they adequately consider functional safety aspects. This dual gap underscores the novelty and relevance of the present study. Furthermore, this work uniquely explores the integration of machine learning with smart traffic management systems to optimize hydrogen fuel cell electric vehicle operations and critically examines the defies and potential solutions for advancing hydrogen refueling infrastructure. The paper concludes by discussing major ongoing challenges and offering perspectives for future research in this transformative field.